nhl glossary // linehype

start_here

elo & the model.

These are the numbers behind the model — our own ratings and how we compare them to the betting market. They work the same in every sport.

ELO
power rating

A single number for how good a team is right now, borrowed from chess (where it ranks players). Every team starts at 1500 — exactly average. Win and your number climbs; lose and it drops. Crucially, beating a great team moves it more than beating a bad one, so the rating quietly accounts for who you’ve played. Above 1500 = better than average. The gap between two teams’ Elo is all you need to estimate who’d win.

1500 = average · 1550+ = strong contender · 1450 = rebuilding
Win Prob
win probability

Our model’s estimate of a team’s chance to win a specific game — calculated from the Elo gap between the two teams, plus a bump for playing at home.
Implied %
market implied probability

The win chance baked into a betting line. A team at −150 is "implied" to win about 60% of the time. It’s the market’s opinion, translated into a percentage so we can compare it to ours.

−150 ≈ 60% · −110 ≈ 52% · +130 ≈ 43%
De-vig
removing the house cut

Betting lines include the house’s built-in profit margin (the "vig"), which makes both sides add up to more than 100%. We strip it out so the market number is a fair, honest probability — apples to apples with our model.
Edge
model vs. market gap

The difference between our model’s win probability and the market’s. A big edge means the math and the money disagree — the game worth a second look. It’s not a betting tip; it’s where two smart estimates part ways.

we flag any gap of 5 percentage points or more

on_the_ice

skaters.

G / A / P
goals, assists, points

The scoring trio. Points (goals + assists) is the headline number for any skater.

70+ points is a star season
+/−
plus-minus

Goal differential while a player is on the ice at even strength — plus when his team scores, minus when it’s scored on.

positive over a season is solid two-way play
SOG
shots on goal

Shots that would go in if not stopped. Volume and shot generation.

250+ in a year is a high-volume shooter
TOI
time on ice

Minutes played per game. The clearest sign of how much a coach trusts a player.

top players log 20+ minutes
PIM
penalty minutes

Time spent in the box. High can mean an enforcer or undisciplined play.

context matters more than the number

between_the_pipes

goaltending.

SV%
save percentage

The share of shots a goalie stops. The first number you check on any netminder.

.905 average · .915+ good · .920+ elite
GAA
goals against average

Goals allowed per 60 minutes. Lower is better, but it’s partly a team-defense stat.

2.50 is solid · under 2.20 excellent
GSAx
goals saved above expected

How many goals a goalie saved beyond what an average goalie would, given the danger of the shots faced. The modern goalie stat.

positive = stealing games
SO
shutouts

Games where a goalie allows zero goals. The rare full-game gem.

5+ in a season is a great year

the_analytics

puck possession.

Hockey’s analytics revolution is built on shot-attempt math — proxies for who actually controls the play. Most are computed from public play-by-play data.

Corsi (CF%)
shot-attempt share

The share of all shot attempts (on net, missed, or blocked) that go your team’s way while you’re on the ice. A stand-in for puck possession — over 50% means you drive play.

50% = even · 53%+ = controls the game
Fenwick (FF%)
unblocked-attempt share

Like Corsi, but it ignores blocked shots — a slightly cleaner read on sustained pressure.

read it the same way as Corsi
xG
expected goals

Adds up the quality of every shot to estimate how many goals a player or team should have, based on shot danger. Separates real chances from harmless perimeter shots.

beating your xG = finishing; trailing it = bad luck or bad hands
PDO
the luck check

On-ice shooting % plus save % — it tends to settle near 100. Way above usually means a hot streak due to cool off; way below, due to bounce back.

~100 is normal · it always regresses

the_one_number

all-in-one value.

WAR / GAR
wins & goals above replacement

Hockey’s newest frontier: models that fold a skater’s entire game — scoring, possession, penalties — into one value number.

we cite these; we don’t compute them
note
why we cite, not compute

These come from independent sites like Evolving-Hockey, each with its own formula. Our Elo ratings and win probabilities we build ourselves; everything else we attribute. See the methodology.